Fault Detection and Fault Tolerance in Robotics

Abstract

Robots are used in inaccessible or hazardous environments in order
\nto alleviate some of the time, cost and risk involved in preparing
\nmen to endure these conditions. In order to perform their
\nexpected tasks, the robots are often quite complex, thus increasing
\ntheir potential for failures. If men must be sent into these
\nenvironments to repair each component failure in the robot, the
\nadvantages of using the robot are quickly lost. Fault tolerant
\nrobots are needed which can effectively cope with failures and
\ncontinue their tasks until repairs can be realistically scheduled.
\nBefore fault tolerant capabilities can be created, methods of detecting
\nand pinpointing failures must be perfected. This paper
\ndevelops a basic fault tree analysis of a robot in order to obtain
\na better understanding of where failures can occur and how they
\ncontribute to other failures in the robot. The resulting failure
\nflow chart can also be used to analyze the resiliency of the robot
\nin the presence of specific faults. By simulating robot failures
\nand fault detection schemes, the problems involved in detecting
\nfailures for robots are explored in more depth. Future work
\nwill extend the analyses done in this paper to enhance Trick, a
\nrobotic simulation testbed, with fault tolerant capabilities in an
\nexpert system package.